NASA Technical Reports Server (NTRS) 20170001236: Aeroelastic Optimization of Generalized Tube and Wing Aircraft Concepts Using HCDstruct Version 2.0

3.4 MB·English

by  NASA Technical Reports Server (NTRS)

Aeroelastic Optimization of Generalized Tube and
Wing Aircraft Concepts using HCDstruct Version 2.0
Jesse R. Quinlan∗ and Frank H. Gern†
NASA Langley Research Center, Hampton, VA, 23681
Major enhancements were made to the Higher-fidelity Conceptual Design and struc-
turaloptimization(HCDstruct)tooldevelopedatNASALangleyResearchCenter(LaRC).
Whereas previous versions were limited to hybrid wing body (HWB) configurations, the
current version of HCDstruct now supports the analysis of generalized tube and wing
(TW) aircraft concepts. Along with significantly enhanced user input options for all air-
craft configurations, these enhancements represent HCDstruct version 2.0. Validation was
performed using a Boeing 737-200 aircraft model, for which primary structure weight esti-
mates agreed well with available data. Additionally, preliminary analysis of the NASA D8
(ND8) aircraft concept was performed, highlighting several new features of the tool.
Nomenclature
AATT = Advanced Air Transport Technology
BDF = Bulk Data File
BWB = Blended Wing Body
CAD = Computer-Aided Design
CFD = Computational Fluid Dynamics
DLM = Doublet Lattice Method
ECFT = Enhanced Correction Factor Technique
FEM = Finite Element Model
FLOPS = Flight Optimization System
HCDstruct = Higher-fidelity Conceptual Design and structural optimization
HWB = Hybrid Wing Body
LaRC = Langley Research Center
MDO = Multidisciplinary Design and Optimization
NASA = National Aeronautics and Space Administration
ND8 = NASA D8
OML = Outer Mold Line
OpenVSP = Open Vehicle Sketch Pad
STL = Stereolithography
SOL = Solution Sequence
TTT = Transformational Tools & Technologies
TW = Tube and Wing
I. Introduction
Sizing conventional aircraft concepts is commonly performed using low-order regression methods with
databases of relevant configuration data, such as those used in the Flight Optimization Performance
System (FLOPS) tool.1 However, these methods are ill-suited for situations where the target concept is
∗AerospaceEngineer,AeronauticsSystemsAnalysisBranch,1NDrydenStreet,andAIAAMember.
†AssistantBranchHead,VehicleAnalysisBranch,1NDrydenStreet,andAIAAMember.
1of13
AmericanInstituteofAeronauticsandAstronautics
unconventional in one or more ways. For example, hybrid wing body (HWB) or blended wing body (BWB)
aircraft concepts are fundamentally unconventional aircraft due to their highly-integrated structures and
streamlined outer mold line (OML). Similarly, recent efforts to structurally size double-bubble fuselage
configurations,2–4 as depicted in Fig. 1, using conventional sizing methods have suggested a need for more
advanced approaches. In order to meet this demand, the Higher-fidelity Conceptual Design and structural
optimization (HCDstruct) tool2,5 was significantly enhanced to include a generalized tube and wing (TW)
airframe sizing capability, and this latest version of the tool is presented in the current paper.
Whereas early development efforts and applica-
tions of HCDstruct were limited to HWB configu-
rations,2,5–9 recent development efforts have been
focused on extending the structural optimization
methodology to TW configurations. While the gen-
eral structural optimization methodology remains
the same, these new applications required signifi-
cantmodificationstothesourcecodetoaccountfor
thefundamentalconfigurationdifferencesassociated
with such concepts. Further, the user controls were
significantly modified to account for the model con-
figuration options available with such concepts.
For the current paper, an overview of the de-
Figure 1. MIT’s D8 aircraft concept featuring a double-
velopment effortsand capabilitiescomprising HCD-
bubble fuselage design (NASA Photo).
struct 2.0 are presented in section II, including pri-
marilyageneralizedTWconfigurationmodelingca-
pability. Applications of HCDstruct 2.0 are then presented in section III, including a validation case using a
Boeing 737-200 model in section III.A and an application to the current NASA D8 (ND8) aircraft concept
being studied under the Advanced Air Transport Technology (AATT) project in section III.B. A summary
of the current work, along with discussion of follow-on efforts, is presented in section IV.
II. HCDstruct Development
HCDstruct was developed at NASA Langley Research Center (LaRC) to fill a critical analysis gap be-
tween high level, lower order approaches commonly used for conceptual design and the low level, detailed,
oftenfinite-element-basedoptimizationapproachescommonlyusedforadvancedpreliminarydesign. Specif-
ically,HCDstructwasdevelopedtocomplementtheFLOPStool,whichisaversatile,multidisciplinarysuite
of computer programs for conceptual and preliminary design and analysis of advanced aircraft concepts.
However, the design sensitivities associated with off-design conditions for the regression-based sizing algo-
rithms1,10 usedbyFLOPSaregenerallyinaccessiblebysuchasimplifiedapproach. Whilethedetailedfinite
element based sizing analyses often performed later in the design cycle, such as those for the HWB by Boe-
ing,11 maytheoreticallyoffersuchinsights, thecomputationalresourcesrequiredfortheseeffortsoftenlimit
them to single-point design analysis. Thus, HCDstruct was developed to bridge the gap between FLOPS’
regression-based sizing techniques and current state-of-the-art finite element based approaches to advanced
preliminarydesigndata. Infact,thetoolhasevolvedtoprovideameansofoptimizingtheprimarystructure
forthefuselageandwingforagivenaircraftconfigurationusingfiniteelementmethodswhileonlyrequiring
FLOPS-level user data.
Since overviews of HCDstruct versions 1.0, 1.1, and 1.2 can be found in Refs. 5, 9, and 2, respectively,
the current section only describes capabilities newly available in version 2.0. With this update, HCDstruct
includes a structural optimization routine for TW aircraft configurations for which the same general sizing
methodologyusedinearlierversionsisemployed. TheHCDstructmethodologyispresentedinFig.2,where
the analysis begins with an OpenVSP12 model of the aircraft concept and a weights schedule comprised of
payload, fuel, andsubsystemsweights. Withthisdata, HCDstructbuildsacompletelyparameterizedfinite-
element model (FEM) of the primary aircraft structure, configures all the load cards for the selected load
cases,andbuildsdoubletlatticemethod(DLM)aerodynamicmodelsforallliftingsurfaces. Theseoutputare
in the form of a complete set of bulk data files (BDFs) that are configured for direct use with NASTRAN13
Solution Sequence (SOL) 200. Execution of SOL 200 results in an optimized primary structural weight
subject to sizing constraints input by the HCDstruct user.
2of13
AmericanInstituteofAeronauticsandAstronautics
The execution of HCDstruct 2.0 assumes a standard naming convention for all input and output files, as
showninFig.3. Fig.3alsoillustratestheexecutionsequence,wheretheinputdecks(*.inp),geometryinput
files(*.stl),andoutputBDFfiles(*.bdf)areASCIIdatafiles,andtheexecutablesHCDstruct andnastran are
theoperating-system-specificcodebuildsforHCDstructandNASTRAN,respectively,locatedinthesystem
path. The HCDstruct executable expects two input parameters, inp dir and out dir, which correspond to
the directories containing the input and output data files, respectively. The nastran executable expects
the hcdstruct exec.bdf BDF file, which contains the executive control input deck for the current case and is
written by HCDstruct. Finally, the output is written to the hcdstruct exec.f06 file directly by NASTRAN.
Figure 2. A notional flowchart detailing the general components of a structural optimization performed using HCD-
struct.
Figure 3. A notional flowchart detailing the general execution process for HCDstruct.
TW applications of HCDstruct utilize the same load cases that were implemented for previous versions
and are described in Ref. 2. These cases include four symmetric loadings (2.5G pull-up, -1.0G push-over,
2P fuselage overpressurization, and 2G taxi bump) and two asymmetric loadings (dynamic overswing and
rudderreversal). DetailsofthestructuraloptimizationproblemusingSOL200andapplicationoftheseload
casesviaSOL144(StaticAeroelasticityAnalysis)subcasesareshowninFig.4. EachsetofBDFfilesoutput
by HCDstruct contains all the data required to execute an instance of NASTRAN SOL 200. The solution
3of13
AmericanInstituteofAeronauticsandAstronautics
sequence is configured to optimize the primary structure by minimizing the total structural weight via the
thickness of the PSHELL elements, and the load cases above are applied via SOL 144 subcases. For every
CQUAD4 element comprising the FEM, an accompanying PSHELL element is configured to represent the
shell properties. The PSHELL elements are linked to a design variable reference card (DVPREL1) and then
toadesignvariable(DESVAR).ForanapplicationtypicalofHCDstruct, theremaybethousandsortensof
thousandsofindependentdesignvariables,andbothstressanddisplacementdesignresponsesareconfigured
as functions of these design variables using the DRESP1 card. Material constraints are applied using the
DCONSTR card. For each iteration of SOL 200, an instance of SOL 144 is spawned for the selected load
cases, a static aeroelastic analysis is performed, and the design responses calculated. SOL 200 then uses a
gradient-based method to modify the PSHELL thickness for each element of the model at each design cycle
subject to the design constraints specified on the DCONSTR cards.
Figure 4. A notional flowchart describing the weight optimization formulation using SOL 200 with HCDstruct.
For TW applications of HCDstruct, the model geometry must be input as two stereolithiography (STL)
files titled wing.stl and fuse.stl corresponding to the wing and fuselage OMLs, respectively. These files
may come from any capable computer-aided design (CAD) tool, but applications to date have utilized the
OpenVSPsoftware. IndependentFEMsofthewingandfuselagecomponentsareconstructedbyHCDstruct
and then automatically connected using glued contact theory13 near the points of intersection with a few
identifyingconfigurationlinesintheHCDstructinputdecks. Theapplicationofgluedcontacttheoryisnew
in version 2.0 and is only supported for TW configurations.
HCDstruct 2.0 offers many configurable features for TW models. Both horizontal and vertical tails can
be configured directly using sweep, cant, and toe angles, and up to two vertical tails can be specified. The
attachmentpointandorientationofthehorizontaltailsurfacearecustomizable, permittingforconventional
empennage assemblies to pi-tail assemblies to T-tail assemblies, and DLM aerodynamic panels are automat-
icallycreatedtomodelallthetailsurfaces. ExamplesofthetailconfigurationssupportedbyHCDstruct2.0
areshowninFig.5. Enginescanbeplacedanywhereonthefuselageorwinginthecurrentrelease, allowing
for the underwing mounting typical of production airliners like the 737-200.
AnewkeycapabilityofHCDstructincludesauser-specifiedparameterreferredtoasfuselage cuspedness,
which allows for the modeling of double-bubble-inspired fuselage concepts. This cuspedness term, referred
to by DBPCT when using the tool, varies from 0.0 to 1.0 and represents the percentage double-bubbledness
of the fuselage cross section. For example, when the cuspedness term is set to 0.0, the fuselage cross section
resembles two concentric circles, and when the figure is set to 1.0, the fuselage cross section is circular
or elliptical, dependent on the fuselage OML. The variation of cross section as a function of DBPCT is
illustratedinFig.6,whereeachimageshowstheleadingcrosssectionalsliceoftheND8fuselage. TheNASA
4of13
AmericanInstituteofAeronauticsandAstronautics
(a) T-tailconfiguration (b) Conventionaltailconfiguration
(c) Pi-tailconfiguration (d) Twintailconfiguration
Figure 5. Horizontal and vertical tail configurations supported by HCDstruct 2.0.
D8 concept uses cuspedness values on the order of 0.5 to model the curvature of the primary structure.
Further,thecurrentreleaseofHCDstructallowsfortheaerodynamicmodelingofthefuselagesectionsusing
slender body elements. Cross sectional slender body properties are computed automatically once the user
requests fuselage aerodynamic modeling. This capability is key to simulating the net lifting effect of the
widened ND8 fuselage concepts, as compared to standard circular cross sectional fuselages.
III. Application Cases
Unlike for previous versions of HCDstruct, which were applicable only to HWB aircraft concepts and for
whichnoas-flown, full-scaleweightsdatawereavailable, theimplementationofageneralizedTWstructural
optimization methodology allows for validation applications using production aircraft data. For this paper,
the first application of HCDstruct 2.0 is a validation study using the Boeing 737-200 aircraft. This analysis
ispresentedinsectionIII.A.ThesecondapplicationofHCDstruct2.0istheND8aircraftconcept,forwhich
preliminary results are presented in section III.B.
III.A. Boeing 737-200
The Boeing 737-200 aircraft was selected for initial validation of HCDstruct 2.0 due to its comparable size
andpassengercapacitytothatoftheND8andalsoduetothegeneralavailabilityofvalidationdataandwide
acceptance in the airline industry since initial production. The Boeing 737-200 represents a conventional
TW aircraft design, with a circular cross sectional fuselage effectively residing over a swept, tapered main
wing, as shown in Fig. 7. The aircraft has a single horizontal stabilizer and a single vertical tail, configured
in a conventional manner.
A geometric model of the Boeing 737-200 was built using OpenVSP with data from several publicly-
available sources.14,15 A three-view depiction of the OpenVSP model is shown in Fig. 8, where several key
dimensions are notated. The fuselage length and wing span are approximately 106.0ft and 94.75ft, respec-
tively. For the purposes of this validation study, only the fuselage, main wing, and horizontal and vertical
stabilizers were geometrically modeled using OpenVSP. The airfoil data was input directly to OpenVSP
using the profile data available in Ref. 16.
5of13
AmericanInstituteofAeronauticsandAstronautics
(a) DBPCT=0.00 (b) DBPCT=0.25
(c) DBPCT=0.50 (d) DBPCT=0.75
(e) DBPCT=1.00
Figure 6. Variation of the ND8 fuselage cross section as a function of DBPCT.
6of13
AmericanInstituteofAeronauticsandAstronautics
III.A.1. Finite Element Model
A three dimensional aeroelastic FEM of the Boe-
ing 737-200 aircraft was developed based on the
OpenVSP model and three view schematic shown
in Fig. 8. The primary airframe FEM is shown
in Fig. 9(a), which includes the fuselage and main
wingcomponentsconnectedusinggluedcontactthe-
ory. The complete aeroelastic FEM is also shown
in Fig. 9(b), which shows the panel-based represen-
tations (CQUAD4) of the fuselage and main wing,
with bulkheads placed at the front and rear of the
fuselage. The front and rear wing spars were posi-
tioned at 12.5% and 62.5% of chord. Two elevators
Figure7. AnimageoftheBoeing737-200aircraftinflight
andonerudderwereconfiguredtopermittheappli-
(NASA Photo).
cation of both the symmetric and asymmetric load-
ing cases. The elevator and rudder hingelines were
positioned at 75% of chord for the horizontal and vertical stabilizers, respectively.
Thehorizontalandverticaltailstructuresweremodeledusingrigidbarelements(RBAR1)atthequarter
chord locations along with point masses (CONM2) to simulate the inertial loadings. The landing gear and
engines were also modeled using rigid bar elements with point masses and are shown in red. The DLM
aerodynamicpanels(CAERO1)areshownforthemainwingandtailsurfaces,andtheslenderbodyelements
(CAERO2) are rendered simply as the yellow line through the center of the fuselage model, which does not
show the slender body elements nor the corresponding interference tube cross sectional properties. The
complete aeroelastic FEM consists of 3824 CQUAD4 elements, 43 RBAR1 elements, 70 CONM2 masses,
642 CAERO1 panels, and 37 CAERO2 elements. The corresponding SOL 200 case included 1912 DESVAR
designvariablesand8DCONSTRdesignconstraintsintheformofvonMisesstressanddisplacementlimits
based on available material properties and a safety factor of 1.5.
Due to the stiffened panel approach employed by HCDstruct, the user must specify the minimum gauge
thicknessforthePSHELLcardsinconjunctionwitheffectivemanufacturabilityormaterialpropertieslimits.
ForthecaseoftheBoeing737-200, forwhichthematerialpropertiesareknownandavailable, theminimum
gauge thicknesses were specified as 0.25in and 0.1in for the fuselage and wing panels, respectively, based on
materials properties and manufacturing limits found in Refs. 17 and 18.
III.A.2. Structural Optimization Results
The Boeing 737-200 aircraft model presented in the previous section was structurally optimized using HCD-
struct2.0subjecttoallthesymmetricandasymmetricloadcasesdescribedinRef.2, andtheresultsofthis
optimizationarecomparedtothoseofas-flownaircraftdatafoundinRef.19. Theweightconvergencehistory
is shown in Fig. 10, demonstrating relative convergence of the total structural weights for both the compo-
nents and composite structure. Further, in Fig. 10, the as-flown aircraft data are presented as horizontal
lines, and the ordinate has been normalized by the as-flown wing structure data. In Fig. 11, the component
and composite structural weights of the optimized aeroelastic FEM are compared to those of the as-flown
aircraft, wherethecomponentweightshavebeennormalizedbytheas-flownaircraftdata. Thefuselageand
wing structural estimates predicted by HCDstruct 2.0 differ from the as-flown data by approximately 5.0%
and 8.9%, respectively. For the total airframe primary structure, HCDstruct predicts the weight to within
approximately 1.6%, thereby supporting the validity of the HCDstruct 2.0 optimization methodology and
the suitability of the selected load cases for the sizing of general TW aircraft concepts.
As was described previously in the discussion of the SOL 200 methodology in Fig. 4, the thicknesses of
the PSHELL elements representing the structural FEM are optimized during the execution of HCDstruct
and NASTRAN. One way to visualize the results of this optimization process is to examine the optimized
thicknesses of the PSHELL elements over the FEM. For the Boeing 737-200 model, this data is shown in
Fig. 12, where the FEM elements are colored by the corresponding PSHELL element thicknesses. In this
figure, the important role of the minimum gauge thickness can be seen clearly in the coloring of the fuselage
structure. In this application, the fuselage is sized predominately by the minimum gauge thickness, with
regions near the front and rear bulkheads being sized by other factors, such as the inertial loadings of the
7of13
AmericanInstituteofAeronauticsandAstronautics
Figure 8. A three-view of the Boeing 737-200 aircraft OpenVSP model used by HCDstruct.
(a) The primary airframe structure FEM for the Boeing (b) The full Boeing 737-200 aeroelastic FEM model, in-
737-200model. cludingtheaerodynamicpanels,theprimarystructure,the
empennage,enginesystems,landinggear,andcontrolsur-
faces.
Figure 9. NASTRAN renderings of the Boeing 737-200 structural FEM and the full aeroelastic FEM.
8of13
AmericanInstituteofAeronauticsandAstronautics
Figure 10. Structural weight convergence history of the optimization design cycles for the Boeing 737-200 aircraft
model.
Figure11. ComparisonsoftheoptimizedstructuralweightsfortheBoeing737-200conceptusingHCDstructtothose
of the as-flown aircraft, normalized to the as-flown weights data.
9of13
AmericanInstituteofAeronauticsandAstronautics
Figure 12. Optimized shell elements colored by PSHELL thickness for the Boeing 737-200 aircraft FEM.
empennage and gear. Additionally, while a substantial portion of the main wing is sized by the respective
minimum gauge, the effects of bending at the root and stress concentrations around the points of engine
attachment result in the thicker panels displayed in these locations.
III.B. NASA D8 Concept
ThelatestiterationoftheND8conceptisshowninFig.13,illustratingthewidenedfuselage,pi-tail,winglets,
and overall vehicle configuration; the wing span, fuselage length, and fuselage width for the current design
areapproximately118ft,106ft,and17ft,respectively. Forongoingmultidisciplinarydesignandoptimiztiaon
(MDO) efforts leveraging FLOPS mission analysis and vehicle sizing methods, HCDstruct 2.0 is used to
develop physics-based estimates for the airframe weights. With these weight estimates, effective corrective
factors may be devised to size the fuselage and wing structures in FLOPS via the FRFU and FRWI cards,
respectively. Inthissection,theND8aeroelasticmodelispresented,followedbyadiscussionofthestructural
weight optimization results.
Using HCDstruct 2.0, an aeroelastic model of the ND8 was constructed consisting of a FEM for the pri-
marystructure,aswellasaDLMaerodynamicmodelforthefuselageandliftingsurfaces. Thisaeroseroelastic
model is presented in Fig. 14, where the CQUAD4 structural elements and CAERO aerodynamic panels are
shaded blue and CONM2 masses, MPC connectors, and RBAR1 rigid bars are shaded red. The FEM is
comprised of 3927 CQUAD4 elements, 83 CONM2 masses, and 59 RBAR1 rigid bars, and the aerodynamic
model is comprised of 642 CAERO1 panels and 37 CAERO2 slender body elements. The model includes
aileron, elevator, and rudder control surfaces, and the DBPCT parameter was set to 0.5.
The corresponding SOL 200 case included 1997 DESVAR design variables and 8 DCONSTR design
constraintsintheformofvonMisesstressanddisplacementlimitsbasedonavailablematerialpropertiesand
asafetyfactorof1.5. Theminimumgaugethicknesseswerespecifiedas0.50inand0.09inforthefuselageand
wing panels, respectively, based on representative composite materials properties and proprietary analysis
for comparable concepts.
Structural weight optimization results are shown in Fig. 15, where the fuselage, wing, and total airframe
structural weights are shown as a function of design iteration. Convergence of these figures occurs within
about ten design cycles, and the results of a FLOPS sizing analysis are also shown on the plot as horizontal
lines for comparison. The fuselage and wing weights converge to approximately 16,900lbs and 11,000lbs,
respectively,foratotalairframestructuralweightof27,900lbs. ComparingthesetotheFLOPSfuselageand
wing weights of 18,300lbs and 10,600lbs, respectively, suggests the use of FLOPS fuselage and wing weight
correction factors of 0.92 and 1.03, respectively.
10of13
AmericanInstituteofAeronauticsandAstronautics